Development and Validation of Predictive Models of Cardiac ... - SciELO

2 downloads 5803 Views 520KB Size Report
HDD and hospitalization due to congestive heart failure increased the risk of death at hazard ... heart failure, primary valvular disease, and incomplete data.
Development and Validation of Predictive Models of Cardiac Mortality and Transplantation in Resynchronization Therapy Eduardo Arrais Rocha1, Francisca Tatiana Moreira Pereira2, José Sebastião Abreu2, José Wellington O. Lima3, Marcelo de Paula Martins Monteiro2, Almino Cavalcante Rocha Neto2, Camilla Viana Arrais Goés1, Ana Gardênia P. Farias2, Carlos Roberto Martins Rodrigues Sobrinho2, Ana Rosa Pinto Quidute2, Maurício Ibrahim Scanavacca1 Instituto do Coração (InCor) – Universidade de São Paulo1, São Paulo, SP; Hospital Universitário – Universidade Federal do Ceará2, CE; Departamento de Saúde Pública – Universidade Estadual do Ceará3, Fortaleza CE – Brazil

Abstract Background: 30-40% of cardiac resynchronization therapy cases do not achieve favorable outcomes. Objective: This study aimed to develop predictive models for the combined endpoint of cardiac death and transplantation (Tx) at different stages of cardiac resynchronization therapy (CRT). Methods: Prospective observational study of 116 patients aged 64.8 ± 11.1 years, 68.1% of whom had functional class (FC) III and 31.9% had ambulatory class IV. Clinical, electrocardiographic and echocardiographic variables were assessed by using Cox regression and Kaplan–Meier curves. Results: The cardiac mortality/Tx rate was 16.3% during the follow-up period of 34.0 ± 17.9 months. Prior to implantation, right ventricular dysfunction (RVD), ejection fraction < 25% and use of high doses of diuretics (HDD) increased the risk of cardiac death and Tx by 3.9-, 4.8-, and 5.9-fold, respectively. In the first year after CRT, RVD, HDD and hospitalization due to congestive heart failure increased the risk of death at hazard ratios of 3.5, 5.3, and 12.5, respectively. In the second year after CRT, RVD and FC III/IV were significant risk factors of mortality in the multivariate Cox model. The accuracy rates of the models were 84.6% at preimplantation, 93% in the first year after CRT, and 90.5% in the second year after CRT. The models were validated by bootstrapping. Conclusion: We developed predictive models of cardiac death and Tx at different stages of CRT based on the analysis of simple and easily obtainable clinical and echocardiographic variables. The models showed good accuracy and adjustment, were validated internally, and are useful in the selection, monitoring and counseling of patients indicated for CRT. (Arq Bras Cardiol. 2015; [online].ahead print, PP.0-0) Keywords: Heart Transplantation / mortality; Heart Failure / physiopathology; Cardiac Resynchronization Therapy; Follow-Up Studies; Pacemaker, Artificial.

Introduction The main international guidelines strongly recommend (class I) cardiac resynchronization therapy (CRT) for patients with congestive heart failure (CHF) and New York Heart Association (NYHA) functional class (FC) II or III or ambulatory class IV when they have intraventricular conduction disturbances and ejection fraction (EF) ≤ 35% while undergoing optimal medical therapy1. However, 30%–40% of CRT cases do not achieve favorable outcomes, which means that these patients undergo surgery with high risks and costs but with no clinical, hemodynamic, or survival benefits2. Thus, multifactorial indexes or scores

Mailing Address: Eduardo Arrais Rocha • Universidade Federal do Ceará. Av. Padre Antônio Tomás, 3535 / 1301, Cocó. Postal Code 60192-120, Fortaleza, CE – Brazil E-mail: [email protected], [email protected] Manuscript received March 15, 2015; reviewed manuscript May 05, 2015; accepted May 06, 2015.

DOI: 10.5935/abc.20150093

need to be developed to more accurately identify survival predictors and treatment responders3,4. Such indexes should involve variables related to mortality reduction, with high rates of sensitivity and specificity. This work aimed to develop predictive models for the combined endpoint of cardiac death and transplantation (Tx) at different stages of CRT.

Methods This prospective observational study evaluated 116 patients with multisite pacemakers implanted consecutively in a tertiary university hospital between January 2008 and March 2013 (Table 1), who had NYHA FC III or ambulatory FC IV (ambulatory outpatients who were taking oral medications), EF ≤ 35%, QRS ≥ 120 ms (left bundle branch block [LBBB] and right bundle branch block [RBBB] with divisional block or pacemaker rhythm), and optimized treatment. The exclusion criteria were severe comorbidities, previous indication for pacemaker implantation, hospitalization for NYHA FC IV heart failure, primary valvular disease, and incomplete data.

Rocha et al. Development of predictive models in CRT

Table 1 – Baseline characteristics and comparison of the results of some variables during the assessment periods Variables Patients Age (years) Sex (male)

Time 1

Time 2

P Value

Time 3

p Value*

116

114

-

92

-

64.8 ± 11.1

-

-

-

-

69.8%

-

-

-

-

25.8 ± 4.1

-

-

-

-

Beta-blockers

88.7%

89.2%

-

91.8%

ACE-inhibitors

97.4%

96%

Furosemide ≥ 80mg/day

31.9%

17%

-

-

-

Dilated cardiomyopathy

59.4%

-

-

-

-

Ischemic cardiomyopathy

29.3%

-

-

-

-

Chagas disease

11.2%

-

-

-

-

BMI

Atrial fibrillation

95.9%

12%

-

-

-

-

CRT-D

54.2%

-

-

-

-

LBBB

71.55%

-

-

-

-

140

< 0.001

-

-

RBBB with divisional block

12%

Pacemaker

16.3%

Posterolateral vein

45.4%

Anterolateral veins

52.5%

Prior QRS width (ms)

160

Number of hospitalizations due to CHF

108

24

< 0.001

16

0.79*

Ejection fraction (median)

29%

33%

< 0.001

35%

0.03*

LVDD (mm)

70

66

< 0.001

65

0.73*

Systolic BP (mm Hg)

115

119.6

< 0.001

121.8

0.84*

Diastolic BP (mm Hg)

70

80

0.07

70

0.34**

FC IIl (NYHA)

68.1%

8.7%

< 0.001

12%

0.07*

FC IV (NYHA)

31.9%

6.1%

< 0.001

7.6%

0.07*

< 0.001

-

0.06*

DD DD Grade I

34.6%

59.2%

-

63.2%

-

DD Grade II

23.7%

27.1%

-

13.9%

-

DD Grade III

29.7%

8.7%

-

16.4%

-

DD Grade IV

11.8%

4.8%

-

5.0%

0.009*

MR

-

-

0.008

-

No MR

3.4%

5.3%

-

15.3%

Mild MR

50.4%

66.0%

-

56.0%

-

Moderate MR

30.4%

18.7%

-

18.6%

-

Severe MR

15.6%

9.8%

-

9.8%

-

RV dysfunction

20.9%

17%

0.62

12%

0.5*

1.1

1.1

-

1.2

-

Creatinine (mg/dL)

Time 1, preimplantation; time 2, 1 year; time 3, 2 years. *Analysis of time 3 in relation to time 2; QRS width, ejection fraction, left ventricular diastolic diameter and blood pressure were variables without normal distribution (median values); BMI: body mass index; ACE: angiotensin-converting enzyme; CRT-D: cardioverter-defibrillator with biventricular pacing; LBBB: left bundle branch block; RBBB: right bundle branch block; CHF: congestive heart failure; LVDD: left ventricular diastolic diameter; BP: blood pressure; FC: functional class (NYHA); DD: diastolic dysfunction; MR: mitral regurgitation; RV: right ventricle.

Arq Bras Cardiol. 2015; [online].ahead print, PP.0-0

Rocha et al. Development of predictive models in CRT

Of the 147 patients who underwent implantation during the study period, only 116 were included in the study for the following reasons: 4 had an EF >35%, 3 had total atrioventricular block, 2 had primary valvular heart disease, 2 had pacemaker infection, 7 had incomplete data, 4 had loss of capture in the left ventricle electrode, 2 did not undergo complete follow-up, 1 had severe comorbidity, 5 were hospitalized for class IV CHF at the time of inclusion, and 1 died of premature respiratory infection. The electrodes of the right ventricle were positioned preferentially in the apical region (84%). The models used in 92, 12, 10, and 2 patients were from St. Jude Medical, Biotronik, Medtronic, and Guidant, respectively. Patients with concomitant indication for an implantable cardioverter-defibrillator (CRT-D group) (54% of the 116 patients) were also included in this study. This indication was for primary prevention in 47 patients and for secondary prevention in 16 patients. Assessments were performed in the preimplantation period (first analysis), at 1 year after implantation (second analysis), and at 2 years after implantation (third analysis) according to a fixed protocol. We analyzed 12 clinical, 8 electrocardiographic, and 7 echocardiographic variables. The clinical variables were age, sex, body mass index, cardiac cachexia, FC, etiology of cardiomyopathy, cardiac vein where the electrode was positioned in the left ventricle, serum creatinine level, systolic and diastolic blood pressures, use of high-dose loop diuretics (≥  80 mg/day of furosemide), and hospitalization due to heart failure. The electrocardiographic variables were: atrial fibrillation; LBBB or RBBB; previous cardiac pacemaker; 1st-degree atrioventricular block; QRS duration; QRS narrowing after implantation; R wave in the V1 lead in patients with LBBB; and QRS axis in the frontal plane after implantation. The echocardiographic variables were: left ventricular (LV) diastolic and systolic diameters; EF computed using Simpson’s method; degree of diastolic dysfunction (DD) from I to IV; degree of mitral regurgitation from I to III; right ventricular dysfunction (RVD); and dyssynchrony. A 12-lead surface electrocardiogram was recorded at the speed of 25 mm/s and amplitude of 10 mm/mV. The longest duration of the QRS measured in one of the leads of the frontal or horizontal plane, which was the lead with the highest value and thus allowed for better evaluation, was taken into account. Cardiac mortality was defined for deaths of end-stage CHF or for sudden death. Echocardiographic parameters The echocardiographic guidelines for the analysis of various echocardiographic parameters were followed, as well as the guidelines for dyssynchrony for the analysis of such parameters5, 6. Three experienced physicians performed the echocardiographic examinations, 72% of which were performed by the same specialist. The examinations were performed using the GE Vivid 7 Ultrasound System (GE Healthcare, Fairfield, CT, USA). The systolic function analysis of the cardiac chambers was performed using Simpson’s method in the two-dimensional mode. Ventricular diameters were obtained on M-mode

Arq Bras Cardiol. 2015; [online].ahead print, PP.0-0

echocardiography, according to the standard guideline5. Right ventricular function was analyzed qualitatively, differentiated between the presence and absence of any degree of dysfunction5. Diastolic dysfunction analysis was conducted by assessing mitral flow (at rest and after a Valsalva maneuver), tissue Doppler images, and flow propagation speed on color M-mode. Results were used to classify DD into four grades (0, absent; I, mild; II, moderate; III, accentuated or with restrictive dysfunction; and IV, severe or with irreversible restrictive dysfunction)7. The degree of mitral regurgitation was assessed as the percentage of the left atrium filling using color Doppler echocardiography. The percentage was less than 20% in mild reflux, and between 20% and 40% in moderate reflux; values above these percentages indicated severe reflux5. In this practical context, the Coanda effect was interpreted as a moderate reflux when restricted to the atrial sidewall and accentuated when it stretched through the upper pole of the left atrium. All patients provided informed consent, and the ethics committee of the hospital approved the study, whose protocol conforms to the ethical guidelines of the declaration of Helsinki. Statistical analysis The categorical variables were presented as frequencies and percentages, whereas the continuous variables were presented as means and standard deviations, or medians. The categorical variables were compared using the McNemar, Stuart–Maxwell, or chi-square test. The Student t test was used to compare the distribution of approximately normal, continuous variables, and the Wilcoxon/Mann–Whitney U test was used for the comparisons of continuous variables without normal distribution. Distributions were considered significantly different if p < 0.05. The univariate relationship between the clinical, electrocardiographic, and echocardiographic variables and the combined endpoint of cardiac mortality and Tx was evaluated by using the Kaplan–Meier survival curve, log-rank test, and Cox regression analysis. Some continuous variables were assessed to determine a cutoff value. Cox multiple regression models were developed in the following analysis times to assess the independent contribution of each of the significant variables in the Cox univariate model: preimplantation (time 1), first year after CRT (time 2), and second year after CRT (time 3). Variables with p  80 mm

2.68

1.00-7.15

0.048

DBP < 60 mmHg

2.63

1.02-6.75

0.044

ACE inhibitors

4.34

0.98-19.17

0.052

MR grade II

2.50

0.89-7.41

0.08

MR grade III

2.80

0.87-9.43

0.08

3.95

1.45-10.74

0.007

0.001

4.85

1.71-13.73

0.003

0.004

5.97

2.15-16.53

0.001

HR: hazard ratio (hazard ratio in the Cox model); CI: confidence interval, P: level of statistical significance; Diuretic ↑: ≥ 80mg of furosemide; SBP: systolic blood pressure; DBP: diastolic blood pressure; FC III / IV: percentage of functional class (FC) III over FC IV; Hospitalization ≥ 1: one or more hospitalizations due to congestive heart failure (CHF); RV: right ventricular; EF: ejection fraction; LVDD: left ventricular diastolic diameter; ACE: angiotensin-converting enzyme; MR: mitral regurgitation.

multivariate analysis, implying a 30-month cardiac event‑free rate of 93%. The combination of two (class B) and three variables (class C) resulted in 30-month cardiac event-free rates of 61% and 0%, respectively. Analysis of the variables at time 2 (first year after CRT) During time 2 (first year after CTR), 13 variables were significant in the Cox univariate regression model. In the Cox multivariate model, RVD, use of HDD, and hospitalization due to CHF were independently related to increased cardiac mortality/Tx rate, with hazard ratios of 3.5, 5.3, and 12.5, respectively. The significant variables in the multivariate model were also significant in the Kaplan–Meier model, when compared by using the log-rank test. The analysis of the model by using the ROC curve showed an AUC of 0.910, with a sensitivity of 76.4%, a specificity of 96.3%, and an accuracy of 93% (Figure 2). From the combinations of these variables, we were able to construct a model with three classes (Table 4). Class A means low risk of cardiac death/Tx, composed by the absence or presence of only one of the significant variables in the multivariate analysis, resulting in a 30-month cardiac event‑free rate of 98%. The combination of two (class B) and three variables (class C) resulted in 30-month cardiac event-free rates of 65% and 0%, respectively (Figure 2). Analysis of the variables at time 3 (second year after CRT) Hospitalizations due to CHF, use of HDD, FC, DD, RVD, EF